2. Observations and data reduction
The spectrophotometric monitoring was conducted at the ESO 3.6 m telescope in service mode. Typically once a month during the visibility period, up to one hour was dedicated to the programme. The instrument was EFOSC1 with 512512 pixels Tektronix CCD until June 1997, and EFOSC2 with a 2K2K Loral/Lesser chip afterwards. The observations always comprised one or several acquisition images, followed by a low-resolution spectrum with the B300 grism through a slit aligned along components A and B. Exposure time was between 15-30 min, spectral resolution was typically Å, and the spectra covered the wavelength range between 3800 Å and 7000 Å. The seeing was between and . A few datasets have been included in this analysis that were obtained before the proper monitoring started. Some were taken with the same configuration as described above; in addition we used the NTT data shown already by Wisotzki et al. (1993) and a set of spectra obtained with the LDS spectrograph at the Nordic Optical Telescope in February 1994.
The CCD frames were reduced in a homogeneous way, largely following standard procedures. The most critical task was to perform an unbiased simultaneous extraction of the two components. We have employed a three-step deblending algorithm, consisting of the following tasks: (1) Two Gaussians, of the same FWHM and with fixed angular separation but otherwise unconstrained, were fitted simultaneously to each spectral resolution element. (2) The variation with wavelength of the resulting parameters FWHM and centroid location was fitted by low-order polynomials. (3) Another double-Gaussian fit was performed, now with only the two amplitudes as free parameters. The algorithm is described in more detail by Lopez et al. (1998). Inspection of the residual maps revealed no significant deviation from this model. The resulting one-dimensional spectra were recorded as MIDAS table files, thus avoiding loss of information due to data rebinning.
Wavelength calibration frames were obtained from comparison lamp spectra, usually based on Helium/Argon lines. The observing conditions were in many cases not photometric, and standard star spectra for flux calibration purposes were obtained in only a few nights. Therefore only a relative flux calibration could be attempted. Two of the pre-monitoring observations were made through a narrow slit, which in one case was not even aligned. We explain below how these measurements could be incorporated into the analysis.
© European Southern Observatory (ESO) 1998
Online publication: October 22, 1998